Musk Is Not Bypassing ASML
Terafab’s interest in free-electron lasers has been framed as a threat to Europe’s lithography champion. The more consequential possibility is different: Musk may be exploring where the EUV machine should end — and where the factory should begin.
Elon Musk needed only three letters to reopen one of semiconductor manufacturing’s most fundamental questions: FEL. His apparent enthusiasm for free-electron lasers quickly became a story about bypassing ASML. But the real disruption may lie somewhere else — in the architecture surrounding the machine.
🟦 Is Musk actually trying to bypass ASML?
There is little public evidence for that conclusion. Quite the opposite. ASML has already incorporated expected demand from Terafab into its capacity planning for 2027 and 2028. CFO Roger Dassen said in July that Terafab was included in the building plans ASML discusses with customers.
That does not tell us what the final Terafab architecture will look like. Nor does it guarantee which ASML systems the project will ultimately buy. But it makes one increasingly common interpretation difficult to sustain:
Musk is building an alternative to ASML.
ASML does not merely produce EUV light. It builds the extraordinarily complex lithography systems that turn that light into patterns on silicon with nanometre-scale accuracy. Change the way the light is generated, and you have not necessarily replaced the machine. You may have changed one layer of it.
🟦 Then what does “FEL FTW” actually tell us?
Not enough to declare a technological revolution. But enough to pay attention. On August 6, Musk responded “FEL FTW” to an X post arguing that Terafab could use free-electron-laser technology for EUV lithography. It was an endorsement of the idea, not a published engineering blueprint. That distinction matters.
ASML’s current EUV systems generate 13.5-nanometre light through laser-produced plasma or LPP. Microscopic droplets of molten tin are targeted by high-powered CO₂ laser pulses, creating a plasma that emits EUV radiation. The process is repeated up to 50,000 times every second.
A free-electron laser approaches the photon problem differently. Instead of creating EUV through tin plasma, accelerated electrons are manipulated to produce intense radiation.
The significance of Musk’s comment therefore lies in what it does — and does not — imply. It suggests interest in another way of generating EUV light. It does not demonstrate that Musk has developed another way of doing everything an ASML lithography system does. Those are two very different propositions.
🟦 Why rethink the light source at all?
Because advanced lithography consumes photons relentlessly. More usable EUV power can translate into greater productivity and future lithography will continue to put pressure on source performance. But this is where the emerging narrative needs an important correction. LPP has not simply reached the end of the road.
In April 2025, ASML demonstrated a 1,000-watt EUV light source and the company explicitly presented the achievement as evidence that its source architecture can continue toward higher power and productivity. That changes the technological contest.
The question is no longer:
Can LPP scale — or must FEL replace it?
ASML has shown that LPP can scale considerably further.
The more interesting question is:
Where should that scaling take place?
ASML’s trajectory is essentially one of continued integration: make the source inside the lithography architecture increasingly powerful, reliable and productive.
The FEL proposition offers a different philosophy. Instead of asking how far the source attached to an individual scanner can be pushed, it asks whether photon generation should remain attached to the individual scanner at all. That is a much more fundamental challenge. Not to the photon. To the architecture.
🟦 What would a free-electron laser actually change?
Potentially, the economics as much as the physics. Free-electron-laser developer xLight describes its approach explicitly as utility-scale light for the whole fab. Its proposed systems would sit outside the clean-room production environment and provide EUV light to multiple scanners simultaneously. The company also claims substantially higher EUV power than conventional sources.
There is no evidence that Terafab will use xLight, nor that Musk intends to adopt its exact architecture. But xLight demonstrates that the underlying concept is not merely social-media speculation. It is being engineered. And the conceptual shift is considerable.
Today, the logic is broadly:
scanner → dedicated EUV source
The alternative becomes:
central EUV infrastructure → multiple scanners
That changes the economic question. Instead of reproducing an extraordinarily complex EUV-generation system around every scanner, photon generation could become shared infrastructure.
If such a system ultimately works reliably and economically at semiconductor-production scale, its attraction would not simply be a cleaner or more powerful light source. It would introduce a different CAPEX and operating model for the fab.
Electricity is infrastructure. Cooling is infrastructure. Process gases are infrastructure. EUV light could begin to be treated in the same way. The factory would no longer simply contain lithography machines. Part of the lithography system would become the factory itself.
🟦 Where would the real friction for ASML begin?
Not necessarily with scanner sales. With architectural control.
Producing EUV photons is only one part of the problem ASML has spent decades solving. Those photons still have to be collected, shaped and projected onto a wafer with extraordinary precision.
EUV cannot be handled with conventional optical lenses. The system depends on highly sophisticated reflective optics, vacuum environments, reticle and wafer positioning, metrology, computational control and extreme system integration. None of that disappears because the light source changes.
That is why the idea that FEL automatically represents an “ASML killer” is misleading. But the opposite conclusion — that it would mean nothing for ASML — would be equally complacent.
ASML has spent decades integrating the EUV source into a tightly controlled lithography architecture. Its San Diego operation, rooted in the acquisition of Cymer, develops the LPP source technology that sits at the heart of today’s EUV generation.
If photon generation eventually moves outside the individual scanner, part of that integrated architecture moves with it. The scanner may remain indispensable. The boundary around the scanner changes. And that matters because industrial power does not come only from selling a critical machine. It also comes from deciding which functions belong inside that machine, how they interact, and which dependencies customers inherit when they adopt the system.
An external FEL would therefore not necessarily challenge ASML where headlines suggest. It could challenge something subtler:
where ASML’s system begins and ends.
That may eventually have consequences for value capture, supplier relationships and technological lock-in. Those consequences are not yet known. But the architectural question is real.
🟦 Why does Terafab make this more consequential?
Because radical architectures become easier to contemplate when the factory does not yet exist.
Existing semiconductor fabs contain decades of accumulated engineering choices. Equipment dimensions, power systems, vibration management, vacuum infrastructure, clean-room architecture and wafer logistics all constrain what can realistically be changed.
A greenfield megafab starts from another position. Instead of adapting a new light source to an existing factory, the factory itself can theoretically be designed around the requirements of the new light source. That is where Terafab changes the equation.
ASML clearly considers the project serious enough to include expected Terafab demand in its longer-term production planning. And Musk’s industrial history makes the architectural question particularly relevant.
The recurring logic behind Tesla and SpaceX has not simply been to substitute one component for another. It has been to reconsider which parts of a production system belong together, which should be vertically integrated and which constraints can be removed by redesigning the system from first principles.
Terafab creates the possibility of applying that logic to semiconductor manufacturing.
Not:
How do we build another EUV scanner?
But:
What should an EUV factory look like if the entire production system were designed again?
That is a considerably bigger question. And for Europe, it is a considerably more important one. Because ASML’s strategic position is not merely the result of possessing one extraordinary piece of technology. It rests on an architecture.
A network of optics, light generation, mechatronics, metrology, software, suppliers, expertise and decades of accumulated integration has turned EUV into an industrial system.
FEL does not invalidate that system. But it raises the possibility that one layer of it could eventually be unbundled. And once layers become unbundled, control can move.
Signify
Terafab is not an attempt to build another ASML. It may become an attempt to redefine what an EUV fab is.
A free-electron laser does not eliminate the nanometre-scale optics, vacuum systems, metrology and mechatronics on which advanced lithography depends. Nor does it prove that ASML’s existing light-source architecture has reached its limit. ASML’s own 1,000-watt LPP demonstration points in the opposite direction.
What FEL offers is a different scaling philosophy. It potentially decouples photon generation from the individual scanner and turns EUV light into shared industrial infrastructure. That is where the real disruption lies.
Technological and industrial power does not shift only when a challenger builds a better machine. Sometimes it shifts when someone redraws the boundaries between the layers.
The real question is not whether Musk can bypass ASML.
It is whether he can change where the ASML machine ends — and where the fab begins.
Image credit
Illustration: Altair Media / OpenAI
Caption
Not bypassing ASML — rethinking where EUV begins. Terafab’s interest in free-electron lasers raises a deeper question: what happens if EUV generation moves from an integrated source inside each lithography system toward shared infrastructure for the entire fab?
